مدل سازی مشکلات رفتاری کودکان در نظام رفاه کودکان: مراقب، جوانان و ادراکات معلم

Using data from two national United States probability samples, the NSCAW Child Protective Services (N = 5501) and Long Term Foster Care (N = 727) samples, this study identifies predictors of child behavioral problems of children referred to child protective services or in long-term foster care. Measurements are presented from three survey waves using raw score behavioral problem scores provided by caregivers, teachers, and youth. The dependent variable is the square root of the raw Total Problems score from the Child Behavior Checklist (caregivers; [Achenbach, T. M. (1991a). Manual for the child behavior checklist 4-18 and 1991 profile. Burlington, VT: Department of Psychiatry, University of Vermont]), Teachers Report Form (teachers; [Achenbach, T. M. (1991b). Manual for the teacher's report form and 1991 profile. Burlington, VT: Department of Psychiatry, University of Vermont]), or Youth Self-Report (youth; [Achenbach, T. M. (1991c). Manual for the youth self-report and 1991 profile. Burlington, VT: Department of Psychiatry, University of Vermont]). From the perspective of caregivers, children in nonkinship foster homes evidenced higher behavioral problems than did those in birth family homes, kinship foster homes, and other living situations. In contrast (in weighted models), from the perspective of teachers, behavior problems of children in kinship foster homes exceeded those of children in nonkinship foster homes. Child gender interacted with source of report with problem levels for girls being considerably lower than those for boys as reported by teachers, somewhat lower as reported by caregivers, and modestly higher as reported by youth. Ethnicity of caregiver also interacted with report source. In particular, the combination of minority ethnicity and teacher report predicted elevated behavioral problems. Lower educational level of the caregiver predicted increased behavioral problems from the perspective of the teacher but not from that of the caregiver. Multilevel models are developed with and without sampling weights. Implications for practice are discussed.